Problem Definition

Currently, AI features within the Atlassian ecosystem—like Rovo Agents, Rovo Chat, and editor-based AI tools (Jira/Confluence)—operate with either individual instruction sets or rely solely on Atlassian's platform-level moderation. There is no centralized mechanism for Organization Administrators to define a global "baseline" of instructions that apply across all AI interactions within their environment.

Even with perfect AUP enforcement, organisations have their own governance requirements - industry-specific constraints, internal policies, and behavioural baselines that are specific to their environment and will never be covered by a generic AUP. 

Suggested Solution

This request is for a site-level or organization-level instruction configuration that sits above all individual AI features. These instructions would be prepended to all interactions and cannot be overridden by individual agent creators or end users.

• Invisible to Creators: To prevent prompt engineering bypasses, the content of these global instructions should not be visible to agent creators or end users.

• Auditability: All changes to the global instruction set must be logged in the Audit Log.

Workaround

n/a